Hello Tom, Thanks for the clarification. I now understand the main goal of the examples. I was confused by the remarks in the example "(other possibilities will be excluded for lack of join clauses)" as this is not a true statement: some possibilities are not shown for lack of join clauses while others are not shown for the sake of simplicity. I think it would be nice to add what you explained somewhere in the text to indicate this is a partial example with the main goal of illustrating join rels that have no linking clauses are not considered by the optimizer; I got the impression that these two examples are to illustrate how DP works in the optimizer.
best regards, Zeyuan On Mon, Apr 7, 2025 at 9:28 PM Tom Lane <t...@sss.pgh.pa.us> wrote: > Zeyuan Hu <ferrishu3...@gmail.com> writes: > > In > https://github.com/postgres/postgres/tree/master/src/backend/optimizer, > > there are two examples on the dynamic programming (DP) algorithm used in > > the optimizer, which I think > > have some inaccuracy: > > You're right that these examples do not consider the effects of > clauses generated by the EquivalenceClass machinery. But I don't > think the exposition would be improved by mentioning that here. > The point of these examples is that we don't consider joining > rels that have no linking clauses at all. > > We could possibly avoid the inaccuracy by making the examples use > some other operators that are not equijoins. But I wonder if that > would not be more confusing rather than less so. > > regards, tom lane >
